All patients (non-hospitalized and hospitalized) had been diagnosed with SARS-CoV-2 infection by PCR testing from nasal swabs or respiratory secretions between March and November 2020, excluding the possibility of infection with a VOC. All patients gave written informed consent before being included in the study.

Serum samples after the second dose of the BNT162b2 mRNA vaccine from healthcare workers were sent to the diagnostic laboratory of the Center for Virology for antibody titer determinations, and anonymized leftover samples were used for our analyses. SARS-CoV-2 infections were ruled out by regular PCR testing of the vaccinees.

Serum samples were collected approximately three weeks after infection/vaccination (“early” samples, median 19 days, range 9-41 days) as well as approximately six months later (“late” samples, median 176 days, range 107-429 days). Median time to sample collection after symptom onset or second vaccination was similar in all cohorts, as shown in Table 1 .

The D614G, Beta and Delta SARS-CoV-2 strains were isolated from nasopharyngeal swabs from COVID-19 patients. Vero E6 cells (ECACC 85020206) were infected and incubated at 37°C until a cytopathic effect occurred. Cell culture supernatant was harvested and the presence of SARS-CoV-2 was confirmed by PCR. The virus isolates were then passaged two more times in Vero E6 cells. All isolates were controlled to be free of other respiratory viruses by PCR as described (19), and were tested negative for mycoplasma contamination by the MycoAlertTM Mycoplasma Detection Kit (Lonza Group Ltd, Basel, Switzerland). The sequences were determined by next generation sequencing and uploaded to the GISAID database. GISAID accession numbers: D614G variant, EPI_ISL_438123/hCoV-19/Austria/CeMM0360/2020 (19); Beta VOC, EPI_ISL_4236051 (mutations D80A, D215G, L242‐, A243‐, L244‐, K417N, E484K, N501Y, D614G, A701V); Delta VOC, EPI_ISL_4172121 (mutations T19R, G142D, E156del/F157, R158G, L452R, T478K, D614G, P681R, R682W, D950). The cell culture-adaptation R682W acquired in the Delta VOC has been shown not to influence virus neutralization (5).

Spike-specific IgG antibodies were quantified in binding antibody units (BAU)/ml using the WANTAI SARS-CoV-2 IgG ELISA kit (Beijing Wantai Biological Pharmacy Ent.). The test is based on the RBD of the original strain isolated in Wuhan and was performed according to the manufacturer’s instructions. Samples were serially diluted until the endpoint in all instances.

The live virus neutralization tests (NT) were performed as previously described (19). Briefly, two-fold serial dilutions of heat-inactivated serum or plasma samples were incubated with 50–100 TCID₅₀ SARS-CoV-2 (D614G strain, Beta or Delta VOCs) for one hour at 37°C before the mixture was added to Vero E6 cell monolayers (starting dilution of samples 1:10). After three days, NT titers were expressed as the reciprocal of the serum dilution required for protection against virus-induced cytopathic effect. NT titers ≥10 were considered positive.

Statistical analysis was performed with GraphPad Prism (Version 9.2.0). For longitudinal analyses, only samples from patients available at both the early and late time points were included. Seven additional early samples were included in the spike sequence analyses ( Figure 6 ).

RBD-ELISA BAU/ml and NT titers were compared with the Kruskal-Wallis test (3 groups, Dunn’s multiple comparisons post hoc test) or Mann-Whitney test (2 groups). In the case of NT titers < 10, a value of 5 was used for these analyses. Fold reductions of antibody levels over time were calculated for each patient individually using the following formulas:

or

This analysis included only patients with a positive test result in all assays at both time points.

Specific neutralization (NT/ELISA ratios) was calculated for samples that were positive in both assays and log-transformed for ANOVA (three groups, Tukey’s multiple comparisons post hoc test)or a paired t-test (two groups). The ranges of NT/ELISA ratios were expressed as maximum-to-minimum factors, which were calculated by dividing the maximum ratio through the minimum ratio for each cohort, time point and virus variant, respectively. Spearman’s rho was applied to evaluate the correlation between the test results (ELISA and NT), to assess the correlation between NT titers against different variants, and the correlation between ELISA, NT or NT/ELISA ratios and the time point of sample collection (days post onset/2^nd vaccination), respectively. The coefficient of determination (R²) was used to assess variability.

The study population consisted of 54 patients infected between March and November 2020 (24 non-hospitalized and 30 hospitalized) and 34 individuals after a two-dose mRNA vaccination (BNT162b2).

SARS-CoV-2 infection was diagnosed by RT-PCR from nasal swabs or respiratory secretions. Non-hospitalized patients showed a mild to moderate course of the disease. Thirteen of 30 (43%) hospitalized patients were treated in intensive care units ( Table 1 ).

Serum samples were taken at two time points. The first sample was collected approximately three weeks post symptom onset or the second vaccine dose (“early” samples, median 19 days, range 9-41 days). Follow-up samples were collected approximately six months later (“late” samples, median 176 days, range 107-429 days). As shown in Table 1 , the sampling time points were similar in all cohorts. Infected individuals were predominantly male, and hospitalized patients were older than non-hospitalized patients.

Non-hospitalized patients yielded the lowest RBD-specific IgG antibody concentrations at both time points (P < 0.0001), whereas hospitalized and vaccinated individuals developed similar amounts of antibodies (P > 0.05, Figure 1 ). Antibody concentrations declined significantly in all cohorts within six months post disease onset or after the second vaccination ( Figure 1 ), with the strongest waning occurring in vaccinated individuals ( Supplementary Table 1 ).

To analyze functional activities of antibodies, we performed NTs with all samples using an isolate from the early pandemic (D614G virus) and the two VOCs Beta (B.1.351) and Delta (B.1.617.2). In line with our ELISA results, non-hospitalized patients showed significantly lower NT titers against all virus strains than hospitalized and vaccinated individuals at both time points, whereas hospitalized individuals developed the highest titers against all variants at each time point ( Figures 2A–C ). At the early time point, no significant difference in NT titers was observed between hospitalized and vaccinated individuals, while six months later hospitalized patients showed significantly higher NT titers than vaccinees ( Figures 2A–C ). For all cohorts, NT titers were highest against the D614G strain at both time points ( Supplementary Figure 1 ), and no significant differences were observed between the Beta and Delta VOCs.

A decline in NT titers over time was observed in all cohorts for the three virus strains ( Figures 2D–F ), which was significant for the vaccinated and hospitalized cohorts. In the case of non-hospitalized individuals, a significant decrease was only detected for the Delta variant ( Figure 2F ; Supplementary Table 1 ).

At both time points, all hospitalized and vaccinated individuals had neutralizing antibodies against the D614G strain, and only one non-hospitalized patient did not develop such antibodies ( Figures 2 , 3 ). In contrast, a substantial number of non-hospitalized patients were negative against the Beta and Delta VOCs at the early time point (37% and 25%, respectively), with a further decline during the six-month period (54% and 50%, respectively; Figure 3A ). In terms of the proportion of samples with detectable neutralizing activity against the Beta and Delta VOCs, the highest percentages of NT-positive samples were from hospitalized patients ( Figure 3B ), with only a few individuals negative in the Beta and Delta VOC NTs at both time points ( Figure 3B ). A similar result was obtained for vaccinees in the NT with the Beta variant ( Figure 3C ), but while only 3% were negative in the Delta VOC NT at the early time point, this proportion increased to 26% over the following six months ( Figure 3C ).

To assess the relationships between ELISA-binding and D614G-neutralizing antibodies ( Figure 4A ) as well as NTs with different SARS-CoV-2 strains ( Figures 4B–D ), we analyzed the respective assay values (BAU/ml, NT titers) by Spearman correlation tests. Overall, significant positive correlations were observed in all instances (p = < 0.0001, r = 0.67-0.95), but a certain variability was observed, especially in the ELISA-NT analysis with the lowest R² values (R² = 0.58 and 0.45, Figure 4A ).

To investigate the specific NT activities of each individual in detail, we compared the ratios between neutralizing and ELISA-binding antibodies for the three groups (specific NT activities) ( Figure 5 ). NT/ELISA ratios were only calculated for samples with a positive result in both assays, which was the case for most samples tested in the D614G NT (only one non-hospitalized patient with a negative NT had to be excluded, see Figures 2 and 3 ). In contrast, a number of samples showed no neutralization of the Beta or Delta VOCs ( Figures 2 , 3 ), especially in the non-hospitalized cohort.

We found no differences in the D614G NT/ELISA ratios between non-hospitalized and hospitalized individuals at both time points ( Figure 5 ), but the ratios after infection were significantly higher than those obtained from vaccinated individuals, indicating either the presence of more strongly neutralizing antibodies and/or a higher proportion of such antibodies. In all cohorts, an improvement of these ratios was observed up to the late time point. Similar results were found with the ratios calculated for the Beta and Delta VOCs ( Supplementary Figure 2 ). In all instances, the maximum-to-minimum factors of the NT/ELISA ratios were higher after infection than after vaccination ( Supplementary Table 2 ), revealing a greater variability within the infected groups.

To analyze the antibody responses in relation to the time points of sample collection, we performed correlation analyses combining the WT-specific antibody levels as well as the NT/ELISA ratios from all cohorts ( Supplementary Figure 3 ). When we plotted the RBD IgG ELISA values and D614G NT titers against the early sampling time points, we saw a moderate negative correlation (Spearman r = -0.4 and -0.35, respectively) ( Supplementary Figures 3A, C ). No such effect was seen for the early NT/ELISA ratios (Spearman r = 0.23) ( Supplementary Figure 3E ) or with any of these parameters at late time points ( Supplementary Figures 3B, D, F ). In the latter instances, we obtained only a very weak or no correlation (r = 0.06 – 0.26). The individual variation, however, was very high for all combinations tested (R² ≤ 0.1).

As we observed more heterogeneous functional antibody responses in infected than in vaccinated individuals ( Figure 5 ; Supplementary Figure 2 ; Supplementary Table 2 ), which could not be explained by the sampling time point ( Supplementary Figures 3E, F ), we hypothesized that mutations in the S sequence of the infecting strain might influence virus neutralization in individual patients. Leftover material from swabs initially used for the diagnosis of SARS-CoV-2 infection was available from 22 patients ( Table 2 ) and was used for sequencing the spike gene of the respective infecting strain. Fifteen of these patients showed neutralizing antibodies against all variants at both time points. Of the remaining seven patients, only early blood samples were available, which were also NT positive against all variants.

One patient was infected with the D614 strain (originally isolated in Wuhan), ten samples had the D614G mutation only, and samples of eleven patients carried one or two additional substitutions in different regions of S ( Figure 6A ) in addition to D614G ( Tables 2 , 3 ) as compared to the reference sequence (see Methods). As shown in Figure 6A , most mutations in the study cohort were located in the NTD. Therefore, we compared the NT/ELISA ratios from the D614G-infected patients (n = 10) with those that had additional mutations in the NTD (n = 8) ( Figure 6B ). Since we had only early samples for five of these patients, an analysis of the late time point was not possible. Patients previously infected with NTD-mutated strains showed a slight reduction in the D614G- and Delta-NT/ELISA ratios compared to patients infected with viruses carrying only the D614G mutation ( Figure 6B ), which was significant in the case of the Beta VOC, suggesting a possible association between S sequence and variance in neutralizing activities.